**3. Results**

#### *3.1. Differential Expression, Heat Map, and Clustering Analysis from the Microarray Study*

Gene expression profiles for the 200 genes and loci that met the cut-off criteria are shown in Figure 1. The majority of these (136) were downregulated in the TM group. Numerical gene expression data, including expression ratios and *p*-values for all 200 genes and loci, are available in Table S2: Differentially Expressed Genes and Loci. The hierarchical clustering analysis based on Pearson correlation distances showed distinct expression patterns for each group. One gene, *SLC6A4*, coding for the serotonin reuptake transporter, whose *p*-value exceeded the chosen ≤0.05 cut-off by a small margin (actual *p* = 0.057), was included because of its known importance in stress-related mechanisms and health. The expression ratios of all genes and loci met the criterion of ≥2.0.

**Figure 1.** Heat map and hierarchical clustering of differentially expressed genes. The figure shows the relative expression of genes across participants, with hierarchical clustering of genes (rows) and participants (columns). Each colored bar indicates the degree to which the Z-score-normalized expression for that gene is either greater than (red) or less than (green) the median value for the gene.

#### *3.2. qPCR Validation of Microarray Data*

Verification and validation of the microarray data were conducted on 15 key genes selected from the 200. Verification of microarray data on these selected genes compared microarray output to results obtained with qPCR (Table 3). Preliminary validation (Table 4) involved larger groups that were less well-matched demographically (see Table 2) than those in the microarray comparison (see Table 1).


**Table 3.** Comparison of Microarray Output with qPCR Output in the Small Groups.

\* TM participant 4 W lacked sufficient RNA for qPCR, so a demographically matched control (40) was dropped from the qPCR analysis, leaving *N* = 10. \*\* ratio of normalized mean expression values, reciprocal in the case of downregulation in TM. \*\*\* "Fold Difference" = 2 (ΔΔCt); Ct = threshold cycle.

**Table 4.** qPCR Determination of Relative Gene Expression (Ct) and "Fold Difference" \* in Larger Groups.



**Table 4.** *Cont.*

\* "Fold Difference" = 2 (ΔΔCt); Ct = threshold cycle.

#### *3.3. Top Networks*

Principal networks that reflect possible causal and functional significance were found by applying IPA to the 200 genes. The network with the highest p-score, along with its top functions and diseases, is shown in Figure 2. The network with the second highest p-score is shown in Figure 3 (Additional networks with lower p-scores are shown in Figures S1–S3: Networks 3–5). Each network includes molecules central to the network (core molecules) and other molecules that affect or are affected by the core molecules. Network 1 (Figure 2), with a p-score of 37 (i.e., *p* = 1 × <sup>10</sup>−37), is strongly related to stress, inflammation, and the defense response. The largest number of connections are to the NF-κB complex and the interferon alpha complex.

Three differentially expressed genes in Network 1—*IL1B* and *TLR4* (upregulated in the TM group) and *SOCS3* (downregulated in the TM group)—are connected not only to the NF-κB complex but also to interferons alpha and gamma, immunoglobulin, and other core molecules. The major themes of this network are cell-to-cell signaling and interaction, hematologic system development and function, and inflammatory response. Key canonical pathways (CPs) are highlighted in the figure.

Core molecules of Network 2 (p-score 28; Figure 3) include the P38 mitogen-activated protein kinases (P38 MAPKs), which are responsive to many stressors and are involved in apoptosis, autophagy, and cell differentiation; AKT, also known as protein kinase B (PKB), important in signaling pathways regulating cell growth, proliferation, differentiation, and survival; IgG, a key protein activating the complement system for eliminating pathogens; and CAV1, a plasma membrane protein important in coupling integrins to the Ras-ERK pathway and promoting cell cycle progression. CAV1 also connects with a central class of histones (Histone H3) involved in the regulation of glucocorticoid signaling and other genes. The major themes of Network 2 are cellular morphology, cell-to-cell signaling and interaction, and hematopoiesis.

**Figure 2.** Network 1 (p-score 37): Cell-to-cell signaling and interaction, hematologic system development and function, and inflammatory response. Nodes without color denote non-significant genes, solid lines denote direct connections, and dotted lines denote indirect connections. Genes color-coded in red are upregulated and those in green are downregulated in the TM group. CP indicates canonical pathways.

**Figure 3.** Network 2 (p-score 28): Cellular morphology, cell-to-cell signaling and interaction, and hematopoiesis. Nodes without color denote non-significant genes, solid lines denote direct connections, and dotted lines denote indirect connections. Genes color-coded in red are upregulated and those in green are downregulated in the TM group. CP indicates canonical pathways.

#### *3.4. Gene Ontological Process Terms*

Analysis of gene ontological process terms enriched in the 200 differentially expressed genes revealed 12 terms that were statistically significant and potentially meaningful in relation to known effects of the TM program (Table 5).


**Table 5.** Gene Ontological Process Terms.

*3.5. Gene Classification Based on Associated Disease*

Groupings of differentially expressed genes based on associated diseases are shown in Table 6. Sixty-two genes were related to hematologic diseases, 26 to coronary artery disease, 34 to diabetes complications, 49 to inflammation, and 64 to CVD. All these disease-related genes were downregulated in the TM group relative to the control group.

**Table 6.** Differentially Expressed Genes Classified According to Associated Diseases (All Downregulated in the TM Group Relative to Control Group).


#### *3.6. Top Genes Upregulated in the Control Group, and Erythropoiesis-Related Genes*

Table 7 shows the top six genes upregulated in the control group, along with four genes known to be related to erythropoiesis. As described in the Discussion, these top six genes are mainly involved in erythrocyte function.

**Table 7.** Top Six Upregulated Genes in the Control Group and Upregulation of Genes Controlling Erythropoiesis.


## **4. Discussion**

These results show first that the gene expression patterns obtained from the microarray analysis of small, demographically well-matched groups differ from each other in a manner consistent with expectations from prior research. Second, the qPCR results examining the relative expression of a sample of 15 key genes in these small groups as well as in larger, less well-matched groups appeared to uphold the trend of the microarray results. This was despite the poor demographic matching in the larger groups. Third, these expression differences appear to have functional confirmation from known stress effects on health and aging.

Several precautions concerning interpretation deserve mention. Because the group size for the validation step was relatively small, conclusions regarding possible TM effects cannot be generalized to larger populations without further confirming studies. A second precaution is that the control groups did not perform an activity that might qualify as a placebo for the twice-daily practice of TM programs for 38 years. Thirdly, it is possible that the two groups compared by microarray may have differed consistently in some unknown or poorly controlled variable.

The first important strength of the study is the finding of 200 genes in the discovery (microarray) component despite the use of comparatively strict inclusion criteria (expression ratio ≥ 2.0 and *p*-value ≤ 0.05). This ratio cut-off is quite high. Only 275 genes and loci (out of a total 16,247 in our study) reached the ratio criterion, while 2041 met the *p*-value criterion.

The second important strength is the existence of characteristic gene expression patterns for the two groups. The patterns are clear enough to be detected visually (Figure 1). More importantly, these patterns were derived objectively using hierarchical clustering analysis of the 200 genes, with data input consisting only of gene name, expression level, and participant signifier. The patterns derived are distinct and likely reflect functional differences rather than random differences in gene expression. These distinct patterns, combined with the random order of blood sampling and processing across both participants and days, argue against a transient stress or stimulus as the possible cause of between-group differences.

The third and greatest strength lies in the fulfillment of predictions based on prior investigations. Many transcriptomic effects of chronic or extreme stress are known [23–25,45], and prior evidence exists for at least partial reversal of some of these effects by mind–body interventions [36], including by other techniques of meditation [32,33]. Furthermore, independent evidence exists showing that the TM program can reverse long-lasting effects of stress such as symptoms of PTSD [11–13], risk factors for CVD [15–18], chronically high levels of stress-related hormones [17,46,47], and low efficiency of energy metabolism [48]. Taken together, these prior studies predict that transcriptomic patterns associated with stress, such as the CTRA and low energy efficiency (see Introduction), should be prevented or reversed after long-term practice of this program.

In the TM group, evidence of prevention or reversal of the CTRA can be seen first in the IPA network analysis of microarray data. Among the 14 genes in Network 1 (see Figure 2) that were downregulated in the TM group relative to the control group, 11 were associated with inflammation in the disease association analysis. On the other hand, among the 10 genes in Network 1 that were upregulated in the TM group, 7 were associated with the defense response in the analysis by gene ontological process term enrichment.

Further indications that the TM group expresses a low-inflammation trait comes from the individual genes. The pro-inflammatory genes in Network 1 were either direct (e.g., *SOCS3*) or indirect (e.g., *ITGB3*) target genes of NF-κB. Expression of suppressor of cytokine signaling 3 (*SOCS3*) is known to correlate directly with pro-inflammatory cytokine levels [49], as are expression levels of integrin genes (*ITGB3*, *ITGB5*, and *ITGA2B)* [50]. A relative downregulation of these and 45 other genes related to inflammatory disease was found in the TM group, consistent with prevention or reversal of the primary, proinflammatory component of the CTRA.

Based on the upregulation of genes related to disease resistance, e.g., those in the defense response and immune system processes categories from the analysis of gene ontological process term enrichment, seven of which also appear in Network 1, the TM group appeared to have enhanced antiviral, antibacterial, and anti-cancer activities, once again opposite to the CTRA pattern. The roles of specific differentially expressed genes further support this conclusion. Examples include five genes that are associated primarily with anti-cancer activity (*CXCL10*, *MICA*, *FPR2*, *CASP5*, and *CASP7)*, three genes that are associated with both anti-cancer and anti-microbial activity (*OAS1*, *ATF3*, and *IFIT3*), and four genes that are associated primarily with the defense response to viruses and bacteria (*CCL4L1*, *IL1B*, *ANKRD22*, and *TLR4*). This finding for genes upregulated in the TM group appears to confirm prevention or reversal of the second component of the CTRA.

This evidence for prevention or reversal of both components of the CTRA expression pattern raises another key point. Although initial evidence for the CTRA came from studies with severely stressed individuals [24,25], the results of the current study indicate a reduction in the CTRA in the TM group compared to healthy controls. This suggests that nominally healthy 65-year-olds carry a substantial load of stress effects, i.e., an allostatic overload [2,51], possibly due to accumulated effects of mild stressors. This is consistent with previous results from studies of short-term and long-term meditation practice indicating that the TM program reduces stress effects well below the level found in the general population [18,46,47].

Other evidence connecting stress effects with inflammation and disease is found in the relationship between genes grouped through ontological process term enrichment and genes grouped through disease association analysis. Using gene ontological term enrichment, 35 differentially expressed genes were classified as related to "response to stress." Of these 35 genes, 27 also were found among the 49 genes classified under inflammation by disease association analysis, and 30 were found among the 64 genes classified under CVD, consistent with known associations between stress response, inflammation, and CVD. Furthermore, consistent with a close association between inflammation and CVD, 47 of the 49 inflammation genes also were found among the CVD-associated genes. Stress response genes were highly represented in the other disease categories as well. All these stress response genes, as well as all the disease-associated genes, were downregulated in the TM group relative to the control group, a direction more likely to be associated with benefits to health.

Another important prediction from prior research concerns evidence for a stressinduced reduction in energy efficiency. Recent articles by Picard, McEwen et al. summarize the critical roles that mitochondrial energy production and other mitochondrial functions play in stress and adaptation [27,31]. As reviewed by Jevning et al. [48], evidence that practice of TM programs increases energy efficiency includes decreased oxygen consumption, decreased respiratory rate, and decreased blood lactate levels. Lactate, a product of glycolysis that is produced in the blood mainly by erythrocytes, is increased during anaerobic metabolism and decreases acutely in erythrocytes during practice of TM [52]. Two observations in the present study provide evidence for higher energy efficiency in these long-term practitioners. Both may center on the role of SOCS3 in mitochondrial energy metabolism.

*SOCS3*, downregulated in the TM group and prominent in Network 1, codes for a chemokine that is important in regulating energy metabolism through inhibitory effects on AMP-dependent protein kinase (AMPK) and leptin [53]. AMPK is central to energy metabolism in mitochondria, thus affecting cellular and whole-body energy levels [54,55]. Increased SOCS3 due to stress and increased inflammatory cytokines is documented to inhibit AMPK, causing insulin resistance in several tissues [56,57]. SOCS3 also can inhibit STAT3 activation [58], providing another possible route for decreased energy efficiency in mitochondria. Mitochondrial STAT3 plays a direct role in maintaining optimal function of the electron transport chain [59]. It is likely that removing inhibitory effects on AMPK by lowering SOCS3 contributes to the improved energy efficiency observed in practitioners of TM programs.

The second observation is related to this and provides confirmatory evidence for more efficient energy production in the TM group. It involves hematologic system development and function, a major theme in the pathway analyses. The top six genes upregulated in the control group are mainly found in erythrocytes and are critical to erythrocyte function. Such large differences in expression of these genes likely reflect the presence of a greater number of reticulocytes (immature erythrocytes) in the control group blood samples. When erythrocyte production is high, reticulocytes, some of which may be large, even nucleated, can enter the bloodstream and contaminate PBMC samples. If mitochondrial oxidative phosphorylation is inefficient, more oxygen is required for a given level of energy production. Even moderate exercise, the level claimed by all study participants, is likely to produce chronic intermittent hypoxia in those with the lowest energy efficiency, and chronic intermittent hypoxia is known to increase erythropoiesis [60].

Supporting the hypothesis that inefficient mitochondrial energy production causes increased erythrocyte production in the control group, both *GATA1* and *GATA2*, master regulators of erythropoiesis [61], were significantly upregulated in this group. The increased expression of *GATA* normally causes an increase in erythropoietin, the direct stimulant of erythropoiesis [61]. Furthermore, *TAL1*, a regulator of erythropoietin receptor sensitivity [62], was significantly increased in the control group, and expression of *EPOR*, coding for the erythropoietin receptor, was increased as well, though not significantly. All these data are consistent with a substantially lower energy efficiency in the control group compared with the TM group.

Another of the many potentially important observations in this study may deserve mention here. Based on the evidence that chronic stress can cause a decrease in telomere length (for review, see [63]), the increased *TAL1* expression found here in the control group is potentially a mediator of reduced telomerase activity. TAL1 inhibits the promotor of hTERT, the catalytic subunit of telomerase [64]. Elevated TAL1, therefore, could cause a reduction in telomerase activity and telomere length.

Predictably, the larger, less well-matched groups of TM and control participants in which 15 key genes were studied by qPCR showed fewer statistically significant expression differences than were found in the microarray comparison of the well-matched groups. Nevertheless, for each of the main areas discussed, differential expression of one or more key genes reached significance. Thus, results for *SOCS3* and *ITGB5* verified an antiinflammatory state; results for *SOCS3* and *AHSP* verified a state of enhanced energy efficiency, and the result for *CXCL10* (tumor suppressor) verified a higher defense response in the meditation group. It is expected that qPCR data on a larger sample of the 200 genes from the discovery step would give greater confirmation of these outcomes.

In an associated study in preparation [65], the transcriptomic data reported here were compared with cortisol and electroencephalographic (EEG) data from these larger groups, along with similar data from younger groups Results of that study tend to confirm the significance of the present findings in relation to proposed anti-stress and anti-aging effects of this meditation program.
